Open Source Workflow Management Systems: A Concise Survey

Management Systems: A Concise

Survey

Ricardo Garcês, Tony de Jesus, Jorge Cardoso*

and Pedro Valente

University of Madeira, Portugal

*SAP Research, Germany

*University of Coimbra, Portugal

A

The use of open source Workflow Management Systems (WfMS) is appealing for organizations due to its low or inexistent cost and its customization capabilities. In this chapter we analyze ten different open source WfMS using a framework that offers decision makers a starting point for selecting a workflow solution. The framework is to be used as a basis for characterizing WfMS based on a set of 22 parameters.

I

Nowadays, many organizations in the commercial, government and non-profit sectors benefit from the use of open source software [1]. Open source software is having a growing impact on the software industry by becoming an important competitor to commercial software [2]. According to [3], the number of suppliers offering workflow management software is estimated to be two hundred. The se-lection of an open source WfMS solution may be quite a difficult and complex un-dertaking. A sound selection requires a complete analysis of the most popular solutions available. Otherwise, it may lead to the choice of an inadequate work-flow product that will not support efficiently the business processes of an organi-zation.

According to [4], the motivations for using and developing open source software are mixed, ranging from philosophical and ethical reasons to pure practical is-sues. Usually, the first perceived advantage of open source models is the fact that the software is made available gratis or at a low cost. But this characteristic is not exclusive to open source software [5]. What really distinguishes open source soft-ware from softsoft-ware available without a fee is the access to the source code and the right to modify it, the right to redistribute the modifications and to improve the code. Each organization has its particularities, so the characteristics of open source solutions allow the customization of open source workflow systems ac-cording to the functioning and the needs of organizations.

This chapter offers an overview and comparison of ten popular open source WfMS using a comprehensive framework for decision makers, providing a starting point to the complex process of selecting an open source WfMS. In fact, this document

is intended to enable managers to better guide, justify and explain their decisions and choices.

W

MS

Several approaches have been proposed to compare information systems and in-formation technologies. They have been provided by prestigious consulting com-panies such as Andersen Worldwide, Ernst & Young, Deloitte & Touche, Coopers & Lybrand, KPMG and Price Waterhouse.

Since workflow technologies have specific characteristics, existing approaches do not address many important perspectives. Therefore, we propose a new and more complete approach. On the one hand, we want to determine what functionalities are provided by WfMS. On the other hand, we also want to evaluate the installa-tion and usage of WfMS, as well as the definiinstalla-tion of workflow processes. For this reason, and as showed in Table 1, we will focus our attention on the compliance of WfMS with the WfMC reference model [6] and on two functional perspectives:

runtime and design time.

Parameters

Process Definition Application (Interface 1) Workflow Client Application (Interface 2) Invoked Applications (Interface 3)

Other Workflow Enactment Services (Interface 4) WfMC Reference Model

Administration and Monitoring tools (Interface 5) Research Scope

Installation Time Documentation Platform Independent

Easiness of Installation and Utilization Web Based

Other Software Required Middleware Platform DBMS Integration Runtime

Transactions Support Process Definition Time Documentation

Easiness of the Process Definition Web Based Organizational Perspective Functional Perspectives Design Time Workflow Language

WfMC reference model

One of the many principles used by the WfMC is the so-called workflow reference model [6]. This model is a general description of the architecture of a workflow management system, in which the main components and the associated inter-faces are described. In the workflow reference model, the tools for constructing and designing workflows are known as process definition applications (Interface 1). Work items are offered to the employees through workflow client applications (Interface 2). By selecting a work item, an employee can begin performing a spe-cific task for a spespe-cific case. When carrying out a task it may be necessary to start an application. All the application software that can be started from the workflow system are known as invoked applications (Interface 3). According to the WfMC reference model, workflow systems may also able interaction with other workflow engines (Interface 4). Workflow tracking, case control and staff management are supported by administration and monitoring tools (Interface 5).

Runtime and design time perspectives

According to [6], at the highest level, all WfMS may be characterized as providing support in three functional areas:

• build time functions are concerned with defining and modeling workflow processes and their activities;

• runtime control functions are concerned with managing workflow processes in an operational environment and sequencing activities;

• runtime interactions are concerned with human users and other applica-tion tools for processing the various activity steps.

As we can see, these three functional areas can be summarized in two core func-tional perspectives: design time (associated with build time functions) and run-time (gathering runrun-time control functions and their interactions).

The runtime perspective, proposed by our framework, is associated with the in-stallation and the testing of the main functionalities of a workflow solution. It is also related to the analysis of the support that is offered by the workflow solution to workflow processes (e.g. support of transactions and treatment of exceptions). The design time perspective, proposed by our framework, is associated with the task of designing a sample workflow process using the process editor. This func-tional perspective is also related to the ability of the process editor to easily, and in a small amount of time, help us to define a relatively complete workflow proc-ess. Given the importance of these two functional perspectives, it only seems natural that they should be analyzed before choosing a workflow solution.

C

We have chosen ten of the most popular and promising open source workflow systems available nowadays. The final set that we will analyze in this chapter is composed of the following WfMS: Bonita, Enhydra Shark, JawFlow, JBoss jBPM, JFolder, JOpera, OpenWFE, RUNA WFE, WfMOpen and YAWL. Before presenting the results of our comparison, it is crucial to clearly identify the environment on which the analysis of the WfMS was made. The installation and test of the work-flow systems was made by two senior students in Computer Science within the scope of their final project. All the WfMS analyzed were installed and tested in a Intel Pentium M 2.00GHz computer with 1 GB memory, 100 GB disk space and running Windows XP. One way to quickly gain a good impression of a workflow management system is to work through a sample process chosen in advance. The sample workflow process used to test the WfMS’ platforms was composed of 15

different tasks with multiple control structures (AND splits/joins and XOR splits/joins) and by nested workflow definitions. It also included 5 different par-ticipants.

C

10

W

MS

The framework specified in this chapter is now considered to compare the ten workflow systems. Table 2 offers an overview of our findings. Most of the systems are not completely compliant to the WfMC reference model. In fact, only Bonita, OpenWFE and YAWL are fully compliant. Moreover, most of the non compliant WfMS do not provide an interface to interact with other workflow enactment ser-vices. All solutions are platform independent. Two systems have been developed within the scope of a research project (YAWL and JOpera). Regarding installation and testing time, we have discovered a wide range of values that go from only 22 minutes, with OpenWFE, to 12 hours and 47 minutes with WfMOpen. One of the most important aspects that influenced the installation time was the documenta-tion provided. We have reached the conclusion that most of the WfMS studied offer enough documentation in order to correctly install and use the system. Regarding the installation easiness, JFolder’s installation was easy, standing out from the all the other workflow systems. On the opposite end, we found Bonita, JawFlow and WfMOpen with a rather complicated installation procedure. We also discovered that only Enhydra Shark and JOpera do not offer a web based ad-ministration environment.

The process definition applications provided by five of the workflow systems ana-lyzed offered mechanisms that allowed designing our sample process without ma-jor constraints. However, RUNA WFE, WfMOpen, Bonita, JFolder and JBoss jBPM process definition applications were quite limited and unpleasant.

The time spent to define our sample process assumes values that vary from al-most 2 hours to approximately 6 hours. YAWL allowed the quickest process defi-nition. Bonita was the one that required most time to design our sample process, 5 hours and 11 minutes.

Regarding the documentation provided, there is little or no documentation avail-able about the process editor of several WfMS analyzed, like: JawFlow, JBoss jBPM, JFolder and WfMOpen. Finally, XPDL is the process definition language most often used by the workflow systems analyzed.

In the following subsections we will discuss each workflow management system, according to our framework, in greater detail.

Bonita

Bonita was developed in 2003 by a team of 14 engineers, of which, Miguel Valdes Faura, Brice Revenant and François Charoy were the project leaders [7]. The cur-rent version is 2.0 and was released in June, 2006. Bonita is a complete workflow system that provides functionalities to handle long-running, user-oriented work-flows and business processes. It allows to dynamically modifying the definition of a running process in order to take into account events that were not planned. This workflow solution also takes benefit from several services that the integration with a J2EE application server provides, such as transactions, role-based authen-tication and connection with external information systems.

1. WfMC reference model. Bonita is fully compliant to the WfMC reference model specification.

2. Runtime perspective. Bonita’s installation and testing took 1 hour and 56

minutes. The documentation provided by the developers is comprehensive, allow-ing us to install the software without facallow-ing any major problem. However, its web-based user interface is unpleasant and not very user friendly. Bonita requires the installation of JDK 1.4, JOnAS Application Server with Tomcat, Jakarta Ant and a DBMS (database management system). It works upon the middleware platform, Java Message Service, in order to exchange data and events. Bonita offers an easy integration with most database management systems. It also supports exceptions treatment and rollback during process execution.

3. Design time perspective. In order to correctly define our sample workflow

process we have spent approximately 5 hours. The graphical editor provided by Bonita is web-based. It is a Java Applet that allows us to design processes by dragging and dropping each activity. We have found a great amount of documen-tation about this process editor, but some details are not clearly explained. For instance, when defining a process, its sub processes should be defined first. For this reason the definition of our sample workflow process was quite complex. The Applet supports the definition of the organizational model, allowing for the specifi-cation of participants and roles. Bonita implements the Workflow Management Coalitions's XPDL (XML Process Definition Language).

Enhydra Shark

Enhydra Shark was developed by Enhydra.org community in 2003. It is an ex-tendable and embeddable Java workflow engine framework completely based on WfMC specifications [8]. Shark can be used as a simple Java library in a servlet, a swing application, or in a J2EE container. The current version of Enhydra Shark is 2.3 which was released in November, 2008.

1. WfMC reference model. Shark is completely conformal to the WfMC reference

model.

2. Runtime perspective. Enhydra Shark’s installation and testing took 6 hours

and 11 minutes. The documentation provided by the developers was quite straight forward, allowing for a relatively simple installation of the software. In order to properly administrate the workflow system, we should use a commercial administration tool. However, this is not mentioned in the documentation, and this application is not available to download in the project’s homepage. This work-flow system does not offer a web based environment. The administration/client application is very user friendly, allowing for a quite easy testing. This workflow system works upon a middleware platform (CORBA). Shark provides an easy in-tegration with most database management systems and offers mechanisms that support the exceptions treatment during a process execution.

3. Design time perspective. It took 2 hours and 24 minutes to define our

sam-ple workflow process. The documentation provided for the workflow process editor is quite comprehensive. This workflow solution provides, by default, a graphical editor very similar to JPEd (used with WfMOpen) called Together Workflow Editor (TWE). It is very practical and easy to use, assuming itself as a complete and in-teresting editor. TWE supports the design of the organizational perspective and the workflow language used is XPDL.

JawFlow

JawFlow was been developed by Vincenzo Marchese in October, 2006 and cur-rently it is in version 3.0. JawFlow is a workflow engine partially conformal to WfMC directives and completely written in Java. It can be customized using ac-tivities written in Java or in any scripting language supported by the Bean Script-ing Framework [9]. To deploy, test and run JawFlow, we have used the JBoss ap-plication server. However, there are no code dependencies to JBoss.

1. WfMC reference model. JawFlow only offers an embedded administration

(in-terface 5) and client application (in(in-terface 2). It does not offer a process definition application (interface 1). This workflow system also does not offer interfaces to invoke other application (interface 3), or to interact with other workflow enact-ment services (interface 4).

2. Runtime perspective. It took 8 hours and 15 minutes in order to correctly

install and test this WfMS. The documentation provided by the developer is very poor making the installation process quite complex. The process administration environment is web-based and relatively easy to use. JawFlow requires JDK 1.5, Jakarta ant, JBoss and a DBMS. This workflow system works upon a middleware platform (Java RMI and CORBA). JawFlow can be integrated with any database management system and offers mechanisms that support error handling during the execution of a workflow process.

3. Design time perspective. This workflow engine does not provide a process

editor. Any editor supporting XPDL can be used. In our case, we have used JPEd (used with WfMOpen). For this reason, the results presented it the Table 2 are identical to the ones that are described in the WfMOpen design perspective. JBoss jBPM

JBoss jBPM is a flexible and extensible workflow management system. The JBoss jBPM’ core component is the plain Java software for managing process definitions and the runtime environment for execution of process instances [10]. Its last re-lease is version 3.2.3.

1. WfMC reference model. JBoss jBPM offers an administration/client

applica-tion. JBoss jBPM is also able to interact with other applications. However, it is not able to interact with other workflow engines. This workflow system also offers a process definition application.

2. Runtime perspective. It took 1 hour and 9 minutes in order to correctly

in-stall and test a working version of jBPM. The documentation provided was com-prehensive. This was the main factor for making the installation and usage of this workflow solution quite simple. Its web based administration/client application is poor in terms of features offered. For this reason, the use of this workflow solution should require the creation and implementation of a customized client and ad-ministration application. jBPM requires the installation of JDK and Eclipse with the JBoss IDE plugins. This workflow system works upon a middleware platform (Java RMI or CORBA). It offers mechanisms that make jBPM portable across the most popular databases and supports an effective treatment of transactions, al-lowing exceptions treatment and rollback during process execution.

3. Design time perspective. To correctly define our sample workflow process,

using the Eclipse-based tooling available for BPEL, we have spent 2 hours and 45 minutes. The lack of documentation about JBoss jBPM’ process editor reflected negatively upon the ease of the process definition. But this was not the only prob-lem faced. In fact, another probprob-lem found was the decision building block

(XOR-split) had to be directly implemented in the code. Because JBoss jBPM uses BPEL in order to define processes, the definition of sub processes is not supported. This results complex workflow diagrams, which are difficult to analyze and under-stand. The definition of our sample process was therefore quite difficult. This process definition editor supports the specification of the organizational perspec-tive. JBoss jBPM supports two process definition languages: jPDL and BPEL. jPDL is a process language to implement business processes and workflows in Java. BPEL provides process orchestration which is the ability to combine web services into a process execution flow.

JFolder

JFolder (also known as PowerFolder) was developed by Gary Steinmetz in 2004 and is in version 1.1. It is a business application development studio and server that uses a XML based language in order to define workflow processes that run within a J2EE environment. Development and administration takes place through a web browser. JFolder contains features like security, persistence, email, file management and data access [11].

1. WfMC reference model. JFolder offers administration and monitoring tools as

well as a workflow client application. However, it is not able to interact with other applications and with other workflow engines. This WfMS also offers a process definition application.

2. Runtime perspective. JFolder installation and testing took 1 hour and 25

minutes. The amount of documentation provided by the developers is sufficient, allowing us to install the software without facing any major problem. Its web-based administration environment is quite unpleasant, becoming very often con-fusing. This fact makes this workflow solution unattractive from an administra-tion point of view. JFolder requires J2EE, Jakarta Ant and JBoss. This workflow system works upon the middleware platform. The documentation does not indi-cate if is possible to integrate JFolder with other than its default DBMS (hsqldb). JFolder offers mechanisms that support error handling during the execution of the workflow process.

3. Design time perspective. It took 4 hours and 25 minutes in order to design

our sample workflow process. A poor documentation is available for this editor. This tool provides a very limited web-based process editor. It is not based on a "drag and drop" idea. This situation makes it harder to add or edit elements of the diagram. There is also no automated mechanism to save the process definition. All these aspects made the design of our workflow process quite hard and com-plex. It does not support the definition of roles and participants (organizational perspective). The JFolder process editor uses a XML-based proprietary language in order to define workflow processes. This language does not support the defini-tion of sub processes.

JOpera

JOpera is built as a collection of plugins for Eclipse. It is a service composition tool that offers a visual language and an execution platform for building workflow processes. It includes a graphical modeling environment, a light-weight execution engine, and also a set of powerful debugging tools which natively supports the iterative nature of service composition. JOpera has a wide range of applications and implications: from rapid development of service-oriented business applica-tions to classical workflow management and business process automation [12]. JOpera plugin for Eclipse 1.9.11 is the latest release of this system.

1. WfMC reference model. JOpera offers an administration and monitoring tool. It is able to interact with other applications. This workflow system also offers a process definition application.

2. Runtime perspective. The JOpera system was developed with research

pur-poses. Its installation and testing took 1 hour and 56 minutes. The comprehen-sive documentation provided by the developers has allowed us to install the soft-ware without facing any major problem. However, the environment offered, based on Eclipse workbench, is not a practical and user friendly management environ-ment. This poor management environment makes this workflow solution unat-tractive from a usage point of view. JOpera requires the installation of Java JDK and Eclipse. JOpera provides integration with the most popular DBMS and sup-ports a simple exception handling model.

3. Design time perspective. In order to correctly define our sample workflow

process we have spent 2 hours and 26 minutes. Enough documentation related with the graphical editor is provided. In spite the fact that the definition of the process is quite simple, JOpera process editor is quite repetitive, making the defi-nition of our sample process longer. Another problem found is that the processes being designed quickly became confusing and it was difficult to analyze/identify the transitions between tasks. This limitation added to the fact that it does not support the organizational perspective allows us to say that this is a very unat-tractive process editor. The workflow language used by JOpera is JOpera visual composition language.

OpenWFE

OpenWFE is an open source workflow engine that has been developed by Lukas Eder and Nicolas Modryzk. It is a complete Business Process Management suite with four components: an engine, a worklist, a client application and a host for automatic agents. It is written in Java, but features access libraries for languages such as Python, Perl, and Ruby, C# (.NET), PHP and Pnuts [13]. OpenWFE is based on a distributed and web-friendly infrastructure. It offers mechanisms that allow persistence, automated form generation and workflow administration. In November 2006, OpenWFE 1.7.2 was released.

1. WfMC reference model. OpenWFE is completely conformal to the WfMC

refer-ence model.

2. Runtime perspective. OpenWFE’ installation and testing took only 22

min-utes mainly due to its intuitive and user friendly environment. The comprehen-sive documentation provided by the developers allows us to install the software without facing any major problems. The only problem found during the installa-tion was that the documentainstalla-tion available mainly described the installainstalla-tion of the system for a Linux operating system. The web-based administration tool and cli-ent application were very user friendly and simple to use. OpenWFE installation requires JDK and JRE in order to work properly. This workflow system works upon a middleware platform (Java RMI). It is also able to be integrated with all of the most important database systems and it supports an effective treatment of transactions, allowing exceptions treatment and rollback during process execu-tion.

3. Design time perspective. Using the graphical editor provided, we have spent

5 hours and 15 minutes in order to correctly define our sample workflow process. Workflow processes are designed in their own XML based language. The lack of documentation of the process editor made this definition process quite long. Dro-flo is a very limited web based process editor. In fact, it is not based on a "drag

and drop" idea. This situation makes it harder to add or edit element of the flow diagram. Another problem found is that in order to save the XML code generated the user has to copy it and then paste it in a text document. In other words, the editor does not have any option to perform this action. This process editor is so unpractical that in most situations it is much easier to define the workflow proc-ess directly using XML. The definition of our sample procproc-ess was, therefore, quite hard. It supports the definition of the organizational perspective.

RUNA WFE

The Runa Consulting Group has released RUNA WFE, an open source work-flow/business process management environment for jBoss jBPM engine. It is an end user solution for business process management, written in Java, which pro-vides a rich web interface containing a work list handler, a process monitor and a form player. It also supports the interaction with external applications [14]. This workflow solution most recent update is RUNA WFE 2.2 (November 2008).

1. WfMC reference model. RUNA WFE offers an administration/workflow client

application and also supports the interaction with other applications. However it is not able to interact with other workflow engines. Moreover, this WfMS offers a process definition application. RUNA WFE is partially conformal to the WfMC model because it does not interact with other workflow enactment services.

2. Runtime perspective. The installation and testing of RUNA WFE took 2 hours

and 20 minutes. The comprehensive documentation provided was sufficient to install and test this system without facing any major problem. The friendly web based administration/client application offered also contributed to an easy test-ing. RUNA WFE requires the installation of JDK. This workflow system works upon the middleware platform. It offers an easy integration with the most popular database management systems and also supports an effective treatment of trans-actions, allowing exceptions treatment and rollback during process execution.

3. Design time perspective. The definition of our sample workflow process,

us-ing RUNA GPD (a process editor for RUNA that sits upon Eclipse workbench), took 3 hour and 57 minutes. The documentation provided was comprehensive, describing several workflow process definition examples. This workflow system does not allow the definition of sub processes, which results in the creation of complex and confusing workflow diagrams. It also requires the direct implemen-tation of the user forms; which may become quite hard for inexperienced users. For these reasons, the definition of our workflow process was quite complicated. It supports the definition of the organizational perspective. The workflow language used by RUNA WFE is jPDL.

WfMOpen

WfMOpen is a J2EE based implementation of a workflow engine. The workflow component is based on a set of Java interfaces that defines API for workflow management facility. It may be used as the core for any process based application implementation and is well suited in providing solutions for business process management related jobs [15]. In May 2008, the most recent update, WfMOpen 2.2, was released.

1. WfMC reference model. WfMOpen is only partially conformal to the WfMC

model, because it does not interact with other workflow enactment services.

2. Runtime perspective. It took 12 hours and 47 minutes in order to correctly

install and test a working version of WfMOpen. The information available in the documentation provided was confusing, dispersed over the document and in

many aspects insufficient. This was the main reason for making the installation and testing of this workflow solution very complex. Moreover, besides the fact that it is poorly documented, the web-based management environment offered is in many aspects quite user unfriendly. WfMOpen requires the installation of JDK and JBoss. This workflow system works upon a middleware platform (Java RMI, CORBA and SOAP). The database integration is achieved only using the default DBMS of this workflow system. It offers build-in solutions for handling exceptions during a process execution.

3. Design time perspective. To correctly define our sample workflow process

using JPEd, we spent 3 hours and 3 minutes. The lack of documentation about JPEd reflected negatively upon the ease of the process definition, making it quite complex. After understanding how it works, JPEd becomes very practical and easy to use. It supports the organizational perspective. WfMOpen uses XPDL with some extensions to define workflow processes.

YAWL

The YAWL system is an open source workflow solution based on the YAWL (Yet Another Workflow Language) language, designed by Wil van der Aalst, Lachlan Aldred, Marlon Dumas and Arthur ter Hofstede, members of the Faculty of Infor-mation Technology of Queensland University of Technology. The project designers developed this new language by taking Petri nets as a starting point and adding mechanisms to allow for a more direct and intuitive support of the workflow pat-terns identified [16]. YAWL provides direct support for all of the workflow patpat-terns and offers mechanisms that allow persistence, automated form generation and workflow administration [17]. YAWL supports the control-flow perspective, data perspective, and is able to interact with web services. The last version of the sys-tem, version 2.2, was released in November 2008.

1. WfMC reference model. YAWL system is completely conformal to the WfMC

reference model specifications.

2. Runtime perspective. The YAWL system was developed for research

pur-poses. YAWL installation and testing took only 49 minutes. The documentation provided by the developers is comprehensive, describing in detail each step of the installation and allowing us to install the software without facing any major prob-lems. In fact, the installation of the software was simple. This workflow system provides a web based administration/client application that is very user friendly and easy to use. In order to work properly, YAWL system installation requires JRE and Apache Tomcat. This system is compatible with a middleware platform: SOAP. The database integration provided does not support some of the most popular DBMS available. It only offers integration with PostrgreSQL as an alterna-tive to Hypersonic. It allows exceptions treatment during process execution.

3. Design time perspective. Using the graphical editor provided, which is not

web-based, we spent 1 hour and 55 minutes in order to correctly define our sam-ple workflow process. The provided documentation related to the editor was com-prehensive. The definition of our sample process was simple. In fact, the process definition editor uses a small set of elements to design a process, simplifying its analysis. It is also based on a “drag and drop” idea. This situation makes it easier to add or to edit elements of the flow diagram. However, one of the major draw-backs of this workflow solution is that it does not support the organizational per-spective. For this reason, we are not able to associate participants or roles to a task.

R

W

Aalst et al. [16] offers a comparison of the functionality of 15 workflow languages based on a set of workflow patterns. We have a different objective since our aim is to evaluate the main features offered, the easiness of the installation and use of WfMS as well as the easiness of the definition of workflow processes. The research on runtime and design time perspectives of workflow systems is very limited. However, these two perspectives have been somewhat and indirectly addressed by academic papers. In [21], Murray offers a case study that analyses the implemen-tation of a commercially available healthcare workflow system in two hospitals’ settings. The framework proposed also includes a parameter with the same aim as our parameter named organizational perspective. It also proposes the parame-ter ease of use of the WfMS which is similar to our parameter easiness of utiliza-tion. The research developed by Stoilova and Stoilov [22] addresses problems re-lated to the assessment and comparison of workflow management systems. The paper proposes an evaluation template composed by eight categories. The func-tional category is composed by some parameters equivalent to the ones that we have used. These parameters are: modeling process definition, workflow client ap-plication, integration with other workflow engines (supported standards) and ad-ministration and monitoring. The paper also proposes another evaluation category:

usability. This category is related with our parameter: easiness of utilization.

R

V

The selection of an adequate workflow system to manage the business processes of an organization is an important and complex decision that depends on several aspects. The decision is significant due to the wide and heterogeneous set of WfMS available, either commercial or open source. The use of open source solu-tions may become very advantageous for organizasolu-tions since source code as well as the right to modify it allows organizations to address specific requirements. Moreover, there are many success cases using this type of software. Nowadays, open source software is used extensively in the industry. The recent acceptance of Linux and the Apache project are excellent examples of this phenomenon. Due to the success of open source solutions, open source workflow systems have, there-fore, become particularly interesting and appealing to IS and IT decision makers. From a set of open source WfMS currently available, we have chosen the most popular and, in our opinion, most interesting WfMS to be analyzed and com-pared. The framework proposed in this chapter for comparing open source WfMS is based on the WfMC reference model and on the runtime and design time per-spective of workflow systems. This chapter offers an important study for industry decision makers by providing a starting point to the complex process of selecting an open source WfMS.

A

This work has been support by Foundation for Science and Technology (FCT), POCTI-219 and FEDER. The support of SAP Research, CEC Dresden, Germany, is also gratefully acknowledged.

R

1. TDG, Open Source Software: Case Studies Examining Its Use. 2003, The Dravis Group.

2. Peeling, N. and J. Satchell, Analysis of the Impact of Open Source Software. 2001, QinetiQ.

3. Aalst, W.M.P.v.d. and K.v. Hee, Workflow management: models, methods, and systems. 1st edition ed. 2002: MIT Press Cambridge, MA, USA.

4. EWLS, Free Software / Open Source: Information Society Opportunities for Europe? 2000, European Working group on Libre Software.

5. Kenwood, C.A., A Business Case Study of Open Source Software. 2001, The MITRE Corporation.

6. Hollingsworth, D., The Workflow Reference Model. 1995, Workflow Manage-ment Coalition.

7. BONITA. BONITA: Workflow Cooperative System. 2007 [cited 21. May 2007]; Available from: http://bonita.objectweb.org.

8. Shark. Enhydra Shark: Java Open Source workflow engine based on XPDL.

2007 [cited 22.05.2007]; Available from:

http://www.enhydra.org/workflow/shark/index.html.

9. jawflow. jawflow: Java Workflow Manager. 2007 [cited 22.05.2007]; Available

from:

https://www-304.ibm.com/jct03004c/servers/solutions/finder/solution/overview.jsp?solu tion_id=soq74085540080014002%7C30.

10. jBPM. JBoss jBPM. 2007 [cited 22.05.2007]; Available from: http://www.jbpm.org/.

11. JFolder. JFolder - Application development and deployment platform. 2006 [cited 22.05.2007]; 1.1 Alpha:[Available from: http://www.powerfolder.org. 12. JOpera. JOpera Project: Process Support for more than Web Services. 2004

[cited 22.05.2007]; Available from: http://www.iks.ethz.ch/jopera.

13. OpenWFE. OpenWFE - open source workflow engine. 2007 [cited 22.05.2007]; Available from: http://www.openwfe.org/.

14. Runa. RUNA WFE. 2007 [cited 22.05.2007]; Available from: http://runawfe.sourceforge.net/.

15. WfMOpen. WfMOpen. 2005 [cited 22.05.2007]; Available from: http://wfmopen.sourceforge.net.

16. Aalst, W.M.P.v.d., et al., Workflow Patterns. Distributed and Parallel Data-bases, 2003. 14(1): p. 5-51.

17. YAWL. YAWL: Yet Another Workflow Language. 2007 [cited 22.05.2007]; Available from: http://yawlfoundation.org/product/index.php.

18. Fitzgerald, B. and T. Kenny. Open Source Software can Improve the Health of the Bank Balance - The Beaumont. in 24th International Conference on Infor-mation Systems (ICIS). 2003. Seattle.

19. Scacchi, W. OpenEC/B: electronic commerce and free/open source software development. in The 5th Workshop on Open Source Software Engineering, part of The 27th International Conference on Software Engineering (ICSE 2005). 2005. St. Louis, Missouri: ACM Press, New York, NY, USA.

20. Gurbani, V.K., A. Garvert, and J.D. Herbsleb, A case study of open source tools and practices in a commercial setting. SIGSOFT Softw. Eng. Notes, 2005.

30(4): p. 1-6.

21. Murray, M. Strategies for the Successful Implementation of Workflow Systems within Healthcare: A Cross Case Comparison. in 36th Annual Hawaii Interna-tional Conference on System Sciences (HICSS'03). 2003: IEEE Computer Soci-ety Washington, DC, USA.

22. Stoilova, K. and T. Stoilo. Comparison of Workflow Software Products. in Inter-national Conference on Computer Systems and Technologies - CompSys-Tech'2006. 2006. Veliko Tarnovo, Bulgaria.

Bonita, 335 Enhydra Shark, 335 framework, 333 JawFlow, 335 JBoss jBPM, 335 JFolder, 335 JOpera, 335

open source software, 333

open source workflow systems, 335 OpenWFE, 335

RUNA WFE, 335

WfMC reference model, 334 WfMOpen, 335

Workflow Management Systems, 333 YAWL, 335